Merge pull request #60 from till-m/fix-validation-checkpoint

mikemccabe210 · web-flow · commit 064acb0655ea · 2025-10-21T10:48:42.000-04:00
fix: stop overwriting `best.pt` every validation
diff --git a/the_well/benchmark/trainer/training.py b/the_well/benchmark/trainer/training.py
@@ -460,6 +460,7 @@ def train(self):
                     self.save_model(
                         epoch, val_loss, os.path.join(self.checkpoint_folder, "best.pt")
                     )
+                    self.best_val_loss = val_loss
             # Check if time for expensive validation - periodic or final
             if epoch % self.rollout_val_frequency == 0 or (epoch == self.max_epoch):
                 logger.info(

Original file line number	Diff line number	Diff line change
`@@ -460,6 +460,7 @@ def train(self):`
`460`	`460`	`self.save_model(`
`461`	`461`	`epoch, val_loss, os.path.join(self.checkpoint_folder, "best.pt")`
`462`	`462`	`)`
	`463`	`+ self.best_val_loss = val_loss`
`463`	`464`	`# Check if time for expensive validation - periodic or final`
`464`	`465`	`if epoch % self.rollout_val_frequency == 0 or (epoch == self.max_epoch):`
`465`	`466`	`logger.info(`